Physics-informed neural network for predicting myocardial mechanics from in vivo physiological measurements
Structural heart diseases (SHDs) are characterized by progressive myocardial remodeling, yet the tissue-level mechanical properties underlying this process cannot be measured directly in vivo. This limitation restricts the ability of conventional physiological measurements to fully characterize myocardial remodeling and disease progression. Here, we present a physics-informed neural network (PINN) framework that integrates constitutive biomechanics with organ-level physiological measurements to simultaneously infer myocardial biomechanical descriptors and predict physically consistent stress responses. By embedding a hyperelastic constitutive model within the learning process, the framework links physiological measurements with tissue-level mechanics while preserving biomechanical consistency. Using right ventricular remodeling in pulmonary hypertension (PH) as a representative model of SHDs, the proposed framework accurately reproduced nonlinear stress-stretch behavior and outperformed conventional data-driven machine learning approaches for stress prediction. Furthermore, incorporation of PINN-derived tissue-level descriptors with organ-level physiological measurements improved disease stratification and enhanced separation between control, mild, and severe PH remodeling states. These findings support physics-informed constitutive learning as an interpretable framework for integrating organ-level physiology with tissue-level biomechanics and establish a foundation for future biomechanics-informed characterization of myocardial remodeling across SHDs.